R Notebook

# install.packages("tidyverse");
# install.packages("rgdal");
library(tidyverse)
require("maps")
library(geosphere)
library(stringr)
library(rgdal)
library(caret)
library(lubridate)
if (!require(ggmap)) { install.packages('ggmap'); require(ggmap) }
path.to.csv <- '../Year_911_Data.csv'
spd.911 <- read.csv(path.to.csv, TRUE)
spd.911$clearance_date_ts = as.POSIXct(strptime(spd.911$Event.Clearance.Date, "%m/%d/%Y %I:%M:%S %p"))
spd.911$clearance_date_date = as.Date(spd.911$clearance_date_ts)
# View(spd.911)
# path to the FOLDER with the .shp file in it. the second param is the name of the .shp file
# seattle <- readOGR(dsn = path.expand("~/documents/INFO370/project-teamname-v2/maps-api-test"), layer = "Seattle_City_Limits")
# usa <- map_data("state")
# data <- merge(usa, spd.911)
# Red Square coordinates
here_long <-  -122.3095
here_lat <- 47.6560
seattle = get_map(location = c(here_long, here_lat), zoom = 13, maptype = 'roadmap')

Map from URL : http://maps.googleapis.com/maps/api/staticmap?center=47.656,-122.3095&zoom=13&size=640x640&scale=2&maptype=roadmap&language=en-EN&sensor=false

freq_by_desc <- table(droplevels(data$Event.Clearance.Description))
# View(freq_by_desc)

ggplot(as.data.frame(freq_by_desc), 
       aes(x = Var1, y = Freq)) +
       geom_bar(stat = 'identity') +# create bar plot
    coord_flip()

#Traffic related calls, suspicious circumstances, and disturbances are the the most significant threats to pedestrations

        

ggmap(seattle) +
  geom_point(data = data, aes(x = Longitude, y = Latitude, group = Event.Clearance.Description, color = Event.Clearance.Description), alpha = 0.5) +
  facet_wrap(~ Event.Clearance.Description) +
  theme(axis.ticks = element_blank(), 
        axis.text.x = element_blank(),
        axis.text.y = element_blank(),
        legend.position = "none"
        )

# selecting just ID and location data
df_loc <- data %>% dplyr::select(CAD.CDW.ID, Longitude, Latitude)

# figuring out number of clusters
wss <- c()
# clusters 1 to 15
for (i in 1:15) {
  wss[i] <- sum(kmeans(df_loc, centers=i)$withinss)
}
plot(1:15, wss, type="b", xlab="Number of Clusters",
  ylab="Within groups sum of squares")

# fitting model
fit <- kmeans(df_loc, 10)
fit$centers # look at cluster sizes and means. want clusters to be about equal size
fit$cluster
cluster.size <- data.frame(1:10, fit$size)
cluster.size

ggplot(data = cluster.size, aes(x = X1.10, y = fit.size)) +
  geom_bar(stat = 'identity')
ggplot()
ggmap(seattle) +
  geom_point(data = as.data.frame(fit$centers), aes(x = Longitude, y = Latitude), alpha = 0.5)
# looking at cluster means
aggregate(df_loc, by=list(fit$cluster), FUN=mean)

df_loc

# adding data back into dataframe 
# df_loc <- df_loc %>% mutate(cluster = fit$cluster) 
data$cluster <- fit$cluster

# View(data)

# timestamp ->  year  month day hour  minute
# sector -> to factor (there are 17 sectors)
# beat -> to factor (there are 3 beats per sector)

# clean the data a bit more
data$event_clearance_ts = as.POSIXct(strptime(data$Event.Clearance.Date, "%m/%d/%Y %I:%M:%S %p"))
data$event_clearance_date = as.Date(data$event_clearance_ts)
data$event_clearance_month = month(ymd_hms(as.character(data$event_clearance_ts)))
data$event_clearance_day = weekdays(data$event_clearance_date)
data$event_clearance_hr = hour(ymd_hms(as.character(data$event_clearance_ts)))
data$event_clearance_mn = minute(ymd_hms(as.character(data$event_clearance_ts)))
data$Initial.Type.Group = factor(data$Initial.Type.Group)
data$Event.Clearance.Group = factor(data$Event.Clearance.Group)
data$Zone.Beat = factor(data$Zone.Beat)
data$District.Sector = factor(data$District.Sector)
data$event_clearance_day = factor(data$event_clearance_day)
data

col.names <- paste(c(
  "Event.Clearance.Code"
  , "cluster"
  , "Census.Tract"
  , "event_clearance_day"
  , "Event.Clearance.Group"
  , "Event.Clearance.SubGroup"
  , "District.Sector"
  , "Zone.Beat"
  #, "event_clearance_ts"
  # ,"Incident.Location"
  , "event_clearance_hr"
  , "event_clearance_mn"
  , "event_clearance_month" 
  , "Hundred.Block.Location"
  ), collapse="|")
cols <- grep(col.names, colnames(data))
cols
# corr_matrix <- cor(data[,cols]) # correlations between all predictor vars
# corr_matrix

# cutoff <- 0.5 # should be higher in practice

# highly_corr <- findCorrelation(corr_matrix, cutoff=cutoff)
# print(colnames(spd.911)[highly_corr]) # age is highly correalted with pregnant

train.data <- select(data, cols)
train.data
# data <- data %>% droplevels()

# grep("Hundred.Block.Location", colnames(train.data), invert = T)

predictors <- grep("Hundred.Block.Location", colnames(train.data), invert = T)
outcome <- grep("Hundred.Block.Location", colnames(train.data))

# train.data[,predictors]
frame <- data.frame(train.data[,predictors])
frame
out.factor <- train.data$Hundred.Block.Location
as.vector(out.factor)


control <- rfeControl(functions = rfFuncs, method="cv", number=10)
results <- rfe(frame, out.factor, sizes = c(1:13), rfeControl = control) # this will take AWHILE...

results
ggplot(results)

# chosen features
predictors(results)